Wiring error electric leakage user high-precision positioning method and device and storage medium

Abstract

The invention discloses a high-precision positioning method and device for a wiring error electric leakage user, and a storage medium, and relates to the technical field of electric leakage detection. The method comprises the following steps: collecting residual current of a tested transformer area and time sequence data of load current of each user in the tested transformer area; based on the time sequence data, calculating the time sequence correlation strength between the load current and the residual current of each user; generating a differentiated penalty coefficient for each user based on the time sequence correlation intensity; constructing a regression model taking the residual current as a dependent variable and the load current of each user as an independent variable, and reconstructing a target loss function of the regression model by taking a penalty coefficient as a coefficient of a regularization term; and through minimizing the target loss function, a weight coefficient corresponding to the load current of each user is estimated, and then the user with a wiring error electric leakage fault is determined. According to the method, the problem of misjudgment caused by normal user interference in a traditional electric leakage user identification method is effectively solved.

Description

Technical Field

[0001] This invention belongs to the field of leakage current detection technology, and in particular relates to a high-precision method, device and storage medium for locating leakage current in users with wiring errors. Background Technology

[0002] In low-voltage distribution networks, leakage faults caused by wiring errors are common due to improper construction, aging lines, and unauthorized connections by users. These faults not only cause energy loss but can also lead to electrical fires, electric shocks, and other safety accidents. Therefore, quickly and accurately identifying and locating users with leakage faults is of significant engineering importance. With the widespread deployment of smart meters, utilizing the user load current time-series data they collect, combined with data-driven methods such as attribution analysis, provides a new technical approach for accurately identifying users with leakage faults due to wiring errors.

[0003] In theory, in a transformer substation experiencing a wiring error and leakage fault, the leakage current of abnormal users directly constitutes a significant portion of the substation's residual current, and the time-series fluctuation characteristics of the two are highly correlated. In contrast, the correlation between the load current of normal users and the substation's residual current is weaker. Therefore, by constructing a regression model between the substation's residual current and each user's load current, and solving for the weighting coefficients corresponding to each user's load current (i.e., their "contribution" to the residual current), it is theoretically possible to distinguish between abnormal and normal users: the weighting coefficients of abnormal users should be significantly larger (close to 1), while the weighting coefficients of normal users should be close to 0.

[0004] However, in practical applications, it has been found that traditional attribution analysis methods based on multiple linear regression have significant limitations. The electricity consumption behavior of normal users in a distribution area is complex and diverse, often affected by factors such as holidays, business trips, and changes in household size, resulting in short-term vacancy or low-power operation. For example, users may turn off the main power switch when away for extended periods, resulting in zero electricity consumption; short-term trips may only leave standby current for devices such as refrigerators and routers, resulting in minimal electricity consumption. According to the principle of current phasor composition, the current of these normal users with low or zero electricity consumption should not be significantly correlated with the large residual current values ​​that appear during abnormal periods. However, in the regression calculation process, because the load current signals of these users are weak and their characteristics are not obvious, traditional methods cannot accurately assess their true contribution. Instead, they are easily assigned large weight coefficients during the fitting process, leading the model to misclassify them as "leaking users." This misclassification phenomenon is particularly prominent in distribution areas with a large number of users and complex electricity consumption behaviors, seriously affecting the accuracy and reliability of fault identification.

[0005] To improve the sparsity and robustness of regression models, regularization methods, such as Lasso regression, have been introduced in existing technologies. By applying L1 norm penalties to all weighted coefficients, Lasso regression tends to compress some coefficients to zero, theoretically enhancing the distinction between abnormal and normal users. However, Lasso regression employs an indiscriminate penalty strategy, applying the same penalty strength to all user variables. While this uniform penalty compresses irrelevant variables (normal users), it may also excessively weaken the coefficients of truly strongly correlated abnormal user variables, leading to missed detections or decreased identification performance in complex scenarios with multiple leakage points.

[0006] Therefore, existing technologies lack a method for locating leakage current users that can effectively distinguish and differentiate user variables with different associated characteristics. In particular, when a large number of normal users are in a special state of low power consumption, existing methods are prone to misjudgment, and a more accurate and robust solution is urgently needed. Summary of the Invention

[0007] To address the aforementioned deficiencies in existing technologies, the present invention aims to provide a high-precision method, device, and storage medium for locating users with wiring errors and leakage current. This addresses the problem that traditional attribution analysis methods (such as multiple linear regression) fail to effectively distinguish the correlation between user current and residual current in a transformer area when there are many normal users with low or zero current, leading to the misclassification of normal users as leakage current users. Furthermore, it overcomes the technical challenge that existing regularization methods (such as Lasso) may over-compress the actual leakage current user signal due to indiscriminate penalty, thus affecting the accuracy and reliability of leakage current user identification in complex scenarios.

[0008] This invention solves the above-mentioned technical problems through the following technical solution: a high-precision method for locating users with wiring errors and leakage current, comprising:

[0009] Collect time-series data of the residual current of the tested transformer area and the load current of each user in the tested transformer area;

[0010] Based on the time-series data, calculate the time-series correlation strength between the load current and the residual current for each user;

[0011] Based on the temporal correlation strength, a differentiated penalty coefficient is generated for each user; wherein, the weaker the temporal correlation strength, the larger the generated penalty coefficient.

[0012] A regression model is constructed with the remaining current as the dependent variable and the load current of each user as the independent variable. The objective loss function of the regression model is reconstructed by using the differentiated penalty coefficient as the coefficient of the regularization term.

[0013] By minimizing the target loss function, the weighting coefficients corresponding to the load current of each user are estimated.

[0014] Based on the weighting coefficient of each user's load current, users with wiring errors and leakage faults are identified.

[0015] This invention objectively quantifies the fluctuation relationship between each user's load current and residual current by calculating the temporal correlation strength. For users with normal electricity consumption and low usage, the temporal correlation strength is inevitably very weak. Based on this, the generated differential penalty coefficient will increase "inversely" (the weaker the correlation, the greater the penalty). In the reconstructed target loss function, a stronger regularization penalty is imposed on these users, thereby mathematically forcing their corresponding weight coefficients to be close to zero. This fundamentally suppresses the influence of these interference signals on the regression model, directly avoiding the misjudgment situation mentioned in the background technology.

[0016] The "differentiation" mechanism of this invention is bidirectional. For users with genuine wiring errors and leakage current, their load current and residual current in the transformer area have a strong temporal correlation, resulting in a high calculated correlation strength and a small penalty coefficient. During the optimization process, the corresponding weight coefficient is subject to a weaker regularization penalty, thus allowing it to be relatively preserved or even amplified. This enables the regression model to more clearly and prominently capture the contribution of genuine leakage current users, enhancing the distinction between abnormal and normal users.

[0017] This invention provides an adaptive penalty mechanism. It does not rely on fixed thresholds or empirical parameters, but rather dynamically adjusts based on the correlation strength driven by the data itself. Regardless of the complexity of user electricity consumption behavior within the distribution area, or how many users are in a low-power state, this penalty mechanism can automatically identify and differentiate its handling, making the regression model solution more stable and the results more reliable. This greatly enhances the practical value of the method of this invention in large-scale, complex low-voltage distribution areas.

[0018] The technical concept of this invention profoundly reflects an understanding of the physical nature of leakage faults (leakage current is a component of residual current, and the two are strongly correlated), and transforms this into a calculable data feature of time-series correlation strength. This physical insight is then embedded into the data model optimization process of differentiated penalties. This integration makes the algorithm not only mathematically effective but also physically reasonable in power system fault analysis, improving the interpretability and engineering credibility of the solution.

[0019] Furthermore, the time-series correlation strength is the Pearson correlation coefficient between the user load current and the residual current.

[0020] This invention concretizes the temporal correlation strength into the Pearson correlation coefficient, providing a correlation metric that is computationally efficient, well-defined, highly interpretable, and easy to implement in engineering. It effectively supports the differentiated penalty mechanism, thereby significantly improving the reliability and practicality of the algorithm while ensuring recognition accuracy.

[0021] The Pearson correlation coefficient directly and intuitively measures the degree of linear correlation between two current sequences in terms of amplitude fluctuations. This aligns closely with the physical intuition that "leakage current leads to a strong correlation," making it easy to understand and verify. Due to the range characteristics of the Pearson correlation coefficient, it can be easily transformed into a differential penalty coefficient through simple calculations. This results in a significant penalty for weakly correlated users (coefficient close to 0) and a slight penalty for strongly correlated users (coefficient close to 1). The penalty gradient is natural and reasonable, effectively amplifying the difference in weight coefficients between normal and abnormal users.

[0022] Furthermore, the penalty coefficient is calculated using the following formula:

[0023] ;

[0024] in, This represents the penalty coefficient corresponding to the j-th user; This represents the Pearson correlation coefficient between the load current and the residual current of the j-th user. c and c both represent preset constants.

[0025] Furthermore, the timing correlation strength is the maximum cross-correlation coefficient between the user load current and the residual current.

[0026] This invention concretizes the temporal correlation strength into the maximum cross-correlation coefficient, which can accurately capture and quantify the time-delayed synchronous fluctuations that may exist between the user load current and the residual current of the transformer area. This allows for a more complete revelation of the essential correlation characteristics of leakage faults, effectively avoiding the possibility of missing strong correlation signals with fixed delays due to simple instantaneous correlation calculations (such as Pearson coefficients). This further improves the robustness and accuracy of identifying abnormal users in complex leakage scenarios.

[0027] Furthermore, the expression for the target loss function is:

[0028] ;

[0029] ;

[0030] in, This represents the objective loss function to be minimized. Let be the vector of weight coefficients to be estimated; This represents the i-th sampled value of the residual current; This represents the weight coefficient of the j-th user; This represents the i-th sampled value of the load current of the j-th user; represents the penalty coefficient corresponding to the j-th user; M represents the number of sampling points; N represents the number of users in the tested area.

[0031] This invention provides a precise and optimizable mathematical model core for the entire localization method by explicitly reconstructing the target loss function into a residual sum of squares with differentiated weighted L1 regularization terms. The weighted L1 regularization term directly implements the "differentiated penalty" mechanism—it applies different degrees of sparsity constraints to the corresponding weight coefficients based on the magnitude of the penalty coefficient, thereby mathematically ensuring that the coefficients of strongly correlated users are preserved or even highlighted, while the coefficients of weakly correlated users are strongly compressed. Ultimately, this solidifies the algorithm's ability to distinguish between normal and abnormal users at the loss function level, ensuring the feasibility and solution stability of the entire technical solution.

[0032] Furthermore, the target loss function is minimized using a numerical optimization algorithm to estimate the weight coefficients.

[0033] This invention introduces a numerical optimization algorithm to minimize the objective loss function, providing a general, efficient, and stable computational framework for solving nonlinear optimization problems containing complex weighted L1 regularization terms. This enables mathematical models with differentiated penalty mechanisms to be reliably solved in engineering practice, thereby transforming theoretical methodological innovation into a practically executable fault diagnosis process, ensuring the accuracy of weight coefficient estimation results and the overall practicality of the algorithm.

[0034] Furthermore, based on the weighting coefficients of each user's load current, users with wiring errors and leakage faults are identified, specifically:

[0035] If the magnitude of the weight coefficient of the j-th user is greater than or equal to the first preset threshold, then the user is determined to be a user with wiring error and leakage.

[0036] If the magnitude of the weight coefficient of the j-th user is less than or equal to the second preset threshold, then the user is determined to be a normal user.

[0037] Wherein, the first preset threshold is greater than the second preset threshold.

[0038] This invention provides a clear, objective, and operable decision boundary for the weight coefficients output by the algorithm by setting first and second preset thresholds and determining the user status accordingly. It maps continuous coefficient amplitudes to discrete fault diagnosis conclusions, thereby achieving a seamless connection from model calculation to engineering judgment. This makes the identification results of users with wiring errors and leakage clear and reliable. Moreover, the thresholds can be adjusted to flexibly adapt to the sensitivity requirements of different transformer substations, greatly improving the practicality and interpretability of the method of this invention.

[0039] Furthermore, the timing data of the residual current and the current of each user load are complex data containing amplitude and phase information.

[0040] This invention employs complex time-series data containing amplitude and phase information, fully preserving the complete phasor characteristics of current signals in AC power distribution systems. This allows the analysis based on time-series correlation strength and subsequent estimation of complex weighting coefficients to accurately reflect the amplitude and phase relationships between currents. Consequently, it more precisely characterizes the physical process of current synthesis and decomposition under leakage faults, significantly improving the accuracy and physical rationality of identifying users with wiring errors and leakage current in complex three-phase imbalance and harmonic scenarios.

[0041] Based on the same concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program or instructions stored in the memory, wherein the processor executes the computer program or instructions to implement the high-precision location method for users with wiring errors and leakage current as described above.

[0042] Based on the same concept, the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implements the high-precision location method for users with wiring errors and leakage current as described above.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] This invention effectively solves the problem of misjudgment caused by interference from normal users (especially users with low or zero power) in traditional leakage current user identification methods by introducing a differentiated penalty mechanism that is inversely correlated with the time-series correlation strength of user-transformer current. It also addresses the problem that indiscriminate regularization methods may weaken the real leakage current signal, thereby significantly improving the accuracy and reliability of leakage current user location. Attached Figure Description

[0045] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only one embodiment of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a flowchart of the high-precision location method for users with wiring errors and leakage current in an embodiment of the present invention;

[0047] Figure 2 This is a bar chart showing the magnitude of the weighting coefficients for leakage current caused by wiring errors in various embodiments of the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0050] like Figure 1 As shown, the high-precision location method for wiring errors and leakage current in users provided in this embodiment of the invention includes the following steps:

[0051] Step S1: Collect time-series data of the residual current of the tested transformer area and the load current of each user in the tested transformer area.

[0052] Using the residual current transformer deployed at the main outgoing line of the low-voltage side of the transformer in the tested area, and the load current acquisition unit installed in each user's meter box, the residual current time sequence data of the tested area and the load current time sequence data of each user are collected synchronously within a set acquisition time period and at a uniform sampling interval.

[0053] The data collection period should cover one or more complete suspected leakage fault cycles or typical power consumption cycles, and its length can be set to several minutes, several hours or several days according to actual needs. The setting of the sampling interval needs to take into account the amount of data, communication load and fault characteristic fidelity, and can usually be selected in the range of seconds to minutes, such as 1 second, 15 seconds or 1 minute.

[0054] The time-series data of residual current and load current of each user are complex data (i.e., phasor data) containing amplitude and phase information. The amplitude unit is amperes, and the phase references the phase of the common voltage of the distribution area. After data acquisition, the data is uploaded to the data server of the distribution area management system for storage and subsequent processing via power line carrier communication, wireless private network, or fiber optic network.

[0055] To ensure data quality and the reliability of subsequent analysis, data verification can be performed simultaneously during the data collection process. This includes checking data continuity, removing obvious outliers, and interpolating or marking missing data.

[0056] Step S2: Based on the time series data, calculate the time series correlation strength between the load current and the remaining current of each user.

[0057] Based on the complex time-series data collected in step S1, the time-series correlation strength between the load current time-series data and the residual current time-series data of each user in the tested area is calculated to quantify the following or consistency of the fluctuation patterns of the two.

[0058] In a preferred embodiment of the present invention, the temporal correlation strength is specifically calculated using the Pearson correlation coefficient. The calculation formula is as follows:

[0059] (1)

[0060] in, represents the Pearson correlation coefficient between the load current and the residual current of the j-th user, and its value is a real number with a range of [0,1]; M represents the number of sampling points of the time series data, and M is greater than the number of users N; This represents the complex value of the residual current at the i-th sampling point (i.e., the sampled value, including amplitude and phase). This represents the complex numerical mean of the time-series data of the residual current in the transformer area. This represents the complex value of the load current of the j-th user at the i-th sampling point; This represents the complex value mean of the time-series load current data for the j-th user.

[0061] In another embodiment of the present invention, the timing correlation strength is the cross-correlation coefficient between the user load current timing data and the transformer area residual current timing data under different time delays. The maximum value (i.e. the maximum cross-correlation coefficient) is taken as the timing correlation strength, which is applicable to scenarios with fixed response delays.

[0062] The calculation results of the time-series correlation strength directly characterize the degree of synchronization between user load current fluctuations and transformer area residual current fluctuations: for users with wiring errors and leakage, their load current directly constitutes part of the residual current, and the fluctuations of the two are highly synchronized, hence the time-series correlation strength is relatively large (usually greater than 0.5 and close to 1); for normal users, especially those with low electricity consumption or in an idle state, their load current fluctuations are unrelated to the transformer area residual current, hence the time-series correlation strength is relatively small (usually less than 0.5 and close to 0). The time-series correlation strength will serve as the direct basis for generating the differentiated penalty coefficient in the next step.

[0063] Step S3: Generate a differentiated penalty coefficient for each user based on the temporal correlation strength.

[0064] Based on the temporal correlation strength of each user calculated in step S2, a corresponding differentiated penalty coefficient is generated for each user. The core principle of this generation process is that the weaker the temporal correlation strength of a user, the larger the penalty coefficient generated, thereby achieving strong compression of weakly correlated variables in subsequent model optimization.

[0065] In a preferred embodiment of the present invention, the penalty coefficient is directly calculated using the following formula:

[0066] (2)

[0067] in, This represents the penalty coefficient corresponding to the j-th user; This indicates the strength of the time-series correlation between the load current and the residual current of the j-th user; Both and 'c' represent preset constants used to adjust the sensitivity of the penalty or avoid division by zero errors. For example, 'c' can be set to 1. The value is 0.01.

[0068] For users with wiring errors and leakage, the timing correlation strength is relatively high (close to 1), the calculated penalty coefficient is relatively small, and the penalty effect is weak in subsequent optimization.

[0069] For normal users, especially those with low battery or idle battery, the temporal correlation strength is very small (close to 0), and the calculated penalty coefficient will be very large (approaching +∞). In subsequent optimization, a strong penalty will be applied to the corresponding weight coefficient.

[0070] The differential penalty coefficient generated in step S3 will be introduced as a key parameter into the target loss function of subsequent reconstruction, thereby achieving "differential treatment" of user variables with different correlation characteristics. This is the core mechanism by which the method of the present invention can overcome the misjudgment caused by normal user interference in traditional methods.

[0071] Step S4: Construct a regression model with residual current as the dependent variable and load current of each user as the independent variable. Reconstruct the target loss function of the regression model by using the differentiated penalty coefficient as the coefficient of the regularization term.

[0072] A multiple linear regression model is established, with the residual current of the transformer area as the dependent variable and the load current of all users as the independent variable. The mathematical expression of the model is:

[0073] (3)

[0074] in, This represents the estimated value of the residual current in the transformer area at the i-th sampling point; This represents the complex value (i.e., the sampled value) of the load current of the j-th user at the i-th sampling point. The complex weighting coefficient of the j-th user is represented, and its physical meaning characterizes the contribution of the user's load current to the residual current and the phase relationship; N represents the number of users in the transformer area. This indicates the background leakage current of the line.

[0075] The residual current timing data can be represented as follows: The weight coefficient vector to be estimated can be represented as: The time-series data of the load current of the j-th user can be represented as: The goal of a regression model is to find an optimal set of regression parameters. This makes the estimated residual current value As close as possible to the actual sampled value .

[0076] The initial target loss function is defined as the sum of squares of the estimation errors over all sampling points (least squares criterion):

[0077] (4)

[0078] To introduce a differentiated penalty mechanism, the differentiated penalty coefficients generated in step S3, corresponding one-to-one with each user, are used as weight coefficients for the regularization term to reconstruct the initial target loss function. The reconstructed target loss function is as follows:

[0079] (5)

[0080] In the reconstructed target loss function:

[0081] First item The second term is the fitting error term, which forces the model estimates to approximate the true values, ensuring the overall accuracy of the model; This is the weighted L1 regularization term. Wherein, This represents the penalty coefficient corresponding to the j-th user. It is a complex weighting coefficient The magnitude of the regularization term. This regularization term penalizes all weight coefficients, but the strength of the penalty varies. Differentiation based on differences:

[0082] for For users with large values ​​(i.e., normal users with weak sequential correlation), this item has a greater impact on them. Apply stronger penalties to force the optimization algorithm to assign its corresponding penalties. The amplitude is compressed to near 0;

[0083] for For users with low values ​​(i.e., users suspected of leakage with strong temporal correlation), this item applies to them. Apply weak penalties, allowing or even encouraging optimization algorithms to apply their corresponding penalties. The amplitude is preserved or amplified (approaching 1).

[0084] By applying this "user-specific" regularization penalty, the reconstructed target loss function, while pursuing overall fitting accuracy, automatically and significantly increases the difference between the final estimated weight coefficient magnitudes of normal and abnormal users, laying a crucial mathematical model foundation for subsequent accurate judgment.

[0085] Step S5: Estimate the weighting coefficients corresponding to the load current of each user by minimizing the target loss function.

[0086] Step S5 aims to solve the objective loss function reconstructed in step S4, which contains a weighted L1 regularization term, to estimate the complex weighting coefficients corresponding to the load currents of each user. This is a constrained convex optimization problem, and its standard form is as follows:

[0087] (6)

[0088] (7)

[0089] The core task is to find an optimal set of complex weight coefficients. This ensures that, under the above magnitude constraint (i.e., formula (7)), the value of the target loss function J is minimized.

[0090] In a preferred embodiment of the present invention, the primal dual interior-point method is used for solving the problem. The implementation process of this algorithm includes the following key steps:

[0091] Problem Transformation: First, the problem of the first term (L1 norm) and the complex field in formula (6) is solved by introducing auxiliary variables (such as...). By converting it into inequality constraints and separating the real and imaginary parts, it can be transformed into a standard quadratic programming or second-order cone programming problem defined on the real number field.

[0092] Constructing the Lagrangian function: A Lagrangian function is constructed for the transformed problem, while a logarithmic barrier function is introduced to handle inequality constraints. (and auxiliary variable constraints), transforming the constrained optimization problem into a series of unconstrained or simply constrained minimization problems.

[0093] Iterative solution: The primal-dual interior-point method is used for iterative solution. In each iteration:

[0094] Original variable update: along the target loss function with respect to the complex weight coefficients The gradient descent direction is determined, and the influence of Lagrange multipliers (dual variables) is considered when updating. The estimated value.

[0095] Dual variable update: Update the Lagrange multipliers associated with the constraints.

[0096] Central parameter adjustment: Adjust the obstacle parameters so that the iteration points gradually approach the optimal solution from inside the feasible region along the "central path".

[0097] Convergence criterion: Calculate the original residual, dual residual, and complementary gap. When all these values ​​are less than the preset tolerance threshold, the algorithm is considered to have converged and the iteration stops.

[0098] Output result: The converged result The value is used as the complex weighting coefficient corresponding to the final estimated load current of each user.

[0099] The primal dual interior-point method is an existing algorithm. In another specific embodiment of the present invention, other numerical optimization algorithms suitable for sparse regularized regression can also be used to solve the problem, such as the iterative weighted least squares method.

[0100] Regardless of the specific algorithm used, the ultimate goal is to satisfy... Under physical constraints, we can efficiently and stably obtain the weight coefficient estimates that minimize the objective loss function J. These complex weighting coefficients are the core criteria used in subsequent steps to distinguish user states.

[0101] Step S6: Based on the weighting coefficient of each user's load current, determine the users with wiring errors and leakage faults.

[0102] Step S6 aims to use the user complex weight coefficients estimated in step S5 as a basis. The system makes a final determination on the user's wiring status. Its core logic is based on the following physical principle: In the case of wiring error and leakage, the load current of the user with leakage will be almost entirely converted into residual current, and its contribution (i.e., the magnitude of the weighting coefficient) should be close to 1; while the contribution of normal users is negligible, and their weighting coefficient should be close to 0.

[0103] The specific determination method is as follows:

[0104] For each user, calculate their estimated complex weighting coefficients. The amplitude (modulus) .

[0105] Two preset thresholds are set: the first preset threshold (leakage current detection threshold) is used to identify users with leakage current, and is usually set in a value range close to 1; the second preset threshold (normal detection threshold) is used to confirm normal users, and is usually set in a value range close to 0, and satisfies that the first preset threshold is greater than the second preset threshold.

[0106] For each user j, based on the magnitude of its complex weighting coefficients Make a judgment:

[0107] User determination of wiring error and leakage: If If the current load is greater than or equal to the first preset threshold, then the j-th user is determined to be a user with wiring errors and leakage current. This determination indicates that the user's load current is highly correlated with the residual current in the transformer area and contributes significantly, which is consistent with the characteristics of a leakage current fault.

[0108] Normal user determination: If If the load current is less than or equal to the second preset threshold, then the j-th user is determined to be a normal user. This determination indicates that the user's load current contributes negligibly to the residual current.

[0109] Intermediate state handling: For a few Users whose current falls between the second and first preset thresholds can be marked as users to be observed or users suspected of having minor leakage. In practical applications, further analysis can be conducted by combining historical data, on-site investigation, or extending the monitoring time.

[0110] In this embodiment, the first preset threshold is 0.9, and the second preset threshold is 0.1. If If the value is ∈[0.9,1.0], it is determined to be a user with wiring errors and leakage; if If the value is ∈[0,0.1], it is determined to be a normal user.

[0111] The setting of the first and second preset thresholds has been verified in simulations to effectively distinguish between leakage current users and normal users (including users with low power consumption), and has a significant degree of differentiation.

[0112] Finally, the judgment results (such as user ID, weight coefficient amplitude, and judgment status) are summarized and output. For users with identified leakage current, alarm information is generated and reported to the transformer area management system, prompting maintenance personnel to conduct on-site verification and fault handling.

[0113] Through step S6, this invention ultimately achieves a complete and automated diagnostic process from raw current data to specific fault location for the user. Based on the weighting coefficients estimated by the differentiated penalty mechanism, combined with clear threshold judgment rules, it can identify users with wiring errors and leakage faults with high accuracy and high reliability, effectively overcoming the problem of misjudging normal users with small current charges by traditional methods.

[0114] To verify the effectiveness and superiority of the proposed high-precision location method for wiring error leakage users based on variable differentiation penalty, a simulation experiment was conducted on a full-scale simulated low-voltage distribution transformer area containing 204 users. This full-scale simulated transformer area includes 204 users of three phases (A, B, and C), which can realistically reflect the load characteristics and topology of the actual transformer area.

[0115] The simulation constructed a complex scenario involving three households experiencing wiring errors and leakage faults. Specifically, users numbered 16 and 22 (phase A) were designated as abnormal users with wiring errors and leakage faults. The remaining 202 households were all normal users, including some special users who were either idle or operating at low power levels, to simulate common differences in electricity consumption behavior in actual transformer substations.

[0116] Taking into account the current measurement performance of smart meters, the computational efficiency under a large user base, and the timeliness requirements for fault identification, the data acquisition parameters are set as follows: the sampling interval is 1 minute (min level), the sampling span is 24 hours (1 day), and the complex time-series data (including amplitude and phase) of the residual current of the transformer area and the load current of each user are collected simultaneously.

[0117] Based on the above parameters, a complete sample set containing M=1440 sampling points (24 hours × 60 minutes) is obtained for subsequent model training and evaluation.

[0118] To highlight the improvements of this invention, a comparative experiment was designed: traditional complex multiple linear regression and the proposed location method based on variable differentiation penalty were used to fit the same sample set, and the weight coefficients for each user were calculated. Figure 2 As shown.

[0119] After fitting the data using traditional complex multivariate linear regression (i.e., least squares method, without regularization or with uniform Lasso regularization), the magnitudes of the weight coefficients for some key users are as follows:

[0120] Abnormal users: User 16 has a weighting coefficient of 0.8993, and User 22 has a weighting coefficient of 0.8664. Both coefficients are significantly higher than those of most normal users, indicating a strong correlation with residual current.

[0121] Normal user (example): The weighting coefficient of user 110 with low battery power is as high as 0.5882.

[0122] Although the coefficient for abnormal users is high, there are also cases where the coefficient is too high among normal users (e.g., user number 110 exceeds 0.5). If a simple threshold segmentation is used (e.g., using 0.5 as the boundary), these normal users are very likely to be misclassified as abnormal users. This indicates that traditional methods are insufficient in resisting interference from normal users with special electricity consumption behaviors and have limited distinguishing ability.

[0123] After fitting the data using the method of this invention (introducing a differential penalty coefficient based on the Pearson correlation coefficient and reconstructing the target loss function into a weighted L1 regularized form), the resulting weight coefficients exhibit a distinctly different distribution:

[0124] Abnormal users: The weighting coefficients of users No. 16 and No. 22 were both accurately estimated to be 1.0 (reaching the theoretical upper limit), indicating that their load current contributed almost entirely to the residual current.

[0125] Normal users: The weighting coefficients of all normal users (202 in total) were significantly compressed to near 0, with the maximum value being only 0.015.

[0126] Depend on Figure 2 As can be seen, the method of this invention achieves precise preservation (or even enhancement) of the coefficients of abnormal users and strong compression of the coefficients of normal users. A near-absolute gap (1.0 vs. <0.015) is formed between the coefficients of abnormal users and normal users, with extremely significant distinguishability.

[0127] Therefore, simulation results show that the method of this invention can accurately identify the designated leakage users (numbers 16 and 22), and their estimated weights reach the theoretical maximum value of 1.0. Compared with traditional methods, this invention completely overcomes the misjudgment problem caused by interference from normal users (especially those with low electricity consumption or those that are not currently active) through a differentiated penalty mechanism. The weight coefficients of normal users are generally suppressed to an extremely low level, thus fundamentally eliminating the risk of misjudgment at the algorithm level. In complex transformer substations containing more than 200 households, this invention can achieve high-precision positioning with only minute-level and day-spanning data, meeting the dual requirements of timeliness and reliability in practical engineering. This embodiment fully verifies the significant advantages of the method described in this invention in improving the accuracy and reliability of identifying leakage users with wiring errors.

[0128] Example 2

[0129] This invention also provides an electronic device, which includes a memory, a processor, and a computer program or instructions stored in the memory. The processor executes the computer program or instructions to implement the high-precision location method for wiring error leakage in this invention.

[0130] Although not shown, the electronic device includes a processor that can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) or loaded from a storage portion into random access memory (RAM). The processor can be a multi-core processor or may contain multiple processors. In some embodiments, the processor may include a general-purpose main processor and one or more specialized coprocessors, such as a central processing unit, graphics processing unit (GPU), neural network processor (NPU), digital signal processor (DSP), etc. Various programs and data required for device operation are also stored in RAM. The processor, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0131] The processor and memory described above are used together to execute programs / instructions stored in the memory. When the program / instructions are executed by the computer, they can implement the methods, steps, or functions described in the above embodiments.

[0132] Although not shown, embodiments of the present invention also provide a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implements the high-precision location method for wiring error leakage users in embodiments of the present invention.

[0133] Readable storage media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0134] The above description only discloses specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or modifications that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A high-precision positioning method for a wiring error leakage user, characterized in that, The positioning method comprises: Collecting time series data of a residual current and load currents of users in a measured station area; Based on the time series data, calculating a time series correlation strength between the load current of each user and the residual current; Based on the time series correlation strength, generating a differentiated penalty coefficient for each user; wherein the weaker the time series correlation strength is, the larger the generated penalty coefficient is; Constructing a regression model with the residual current as the dependent variable and the load currents of users as the independent variables, and reconstructing an objective loss function of the regression model by taking the differentiated penalty coefficient as the coefficient of a regularization term; By minimizing the objective loss function, the weight coefficients corresponding to the load currents of users are estimated; According to the weight coefficient size of the load current of each user, the user with a wiring error leakage fault is determined.

2. The method according to claim 1, wherein, The time series correlation strength is the Pearson correlation coefficient between the load current of the user and the residual current.

3. The method according to claim 2, wherein, The penalty coefficient is calculated by the following formula: ； wherein, represents a penalty coefficient corresponding to the jth user; represents a Pearson correlation coefficient between the jth user's load current and the residual current; , c all represent preset constants.

4. The method of claim 1, wherein, The time series correlation strength is the maximum cross-correlation coefficient between the load current of the user and the residual current.

5. The method of claim 1, wherein, The expression of the objective loss function is: ； ； wherein, represents a target loss function to be minimized, is a weight coefficient vector to be estimated; represents the i-th sample value of the residual current; represents the weight coefficient of the j-th user; represents the i-th sample value of the load current of the j-th user; represents the penalty coefficient corresponding to the j-th user; M represents the number of sampling points; and N represents the number of users in the measured area.

6. The method of claim 1, wherein, The objective loss function is minimized by a numerical optimization algorithm to estimate the weight coefficients.

7. The method of claim 1, wherein, According to the weight coefficient size of the load current of each user, the user with a wiring error leakage fault is determined, specifically: If the weight coefficient amplitude of the jth user is greater than or equal to a first preset threshold, the user is determined to be a wiring error leakage user; If the weight coefficient amplitude of the jth user is less than or equal to a second preset threshold, the user is determined to be a normal user; Wherein, the first preset threshold is greater than the second preset threshold.

8. The method according to any one of claims 1-7, wherein, The time series data of the residual current and the load currents of users are complex data containing amplitude and phase information.

9. An electronic device comprising a memory, a processor, and a computer program or instructions stored on the memory, wherein the computer program or instructions, when executed by the processor, cause the electronic device to perform the method of any one of claims 1-8. The processor executes the computer program or instructions to implement the wiring error leakage user high-precision positioning method of any one of claims 1-8.

10. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions are executed by the processor to implement the wiring error leakage user high-precision positioning method of any one of claims 1-8.

Images